Transformation

Part one of a promised "Refactoring Trilogy" whose sequels never turned up (nothing else in this archive answers to the name), written for Steve's internal Google blog in late 2005 and republished on the O'Reilly Ruby blog the following spring. The provocation was specific: Java programmers kept telling him they couldn't even try Ruby because it lacked automated refactoring tools. Rather than argue, he tells a conversion story. He had dodged Fowler's book for years because it shipped in the UML series he considered weenie territory; then one winter day in 2002 he opened it in a bookstore, hit an argument that his cherished habit of caching intermediate values in local variables was Evil, and plopped into a squashy armchair to find out whether the author was "actually insane, or merely an idiot." The next day he casually polled twenty developers at work. One had read the book.

It is Ruby advocacy wearing a book review, and the machinery shows in places: the aside rating Ruby a productivity step-function "at least as great as Java was over C++" is pure 2005 fever, and Steve gives the strongest counterargument — code has to change under shifting requirements no matter who wrote it — exactly one paragraph before brushing past it. The construction is the fun part. The essay repeats its own opening word for word halfway through, drifts into a reverie about earth-movers and the little segmented bug the Caterpillar company is named for, and settles the whole dispute in a closing sentence four words long. The needling lands hardest on the book's own fans ("C'mon now. Admit it. You skimmed."), the people who love a tool catalogue without ever absorbing why the tools were needed.